(;GM[1]FF[3]
RU[Japanese]
SO[]
EV[Computer go 9x9 ladder]
BR[]WR[14k*]CP[]SZ[9]HA[0]KM[5.5]
PW[Gnu go 3.0.0]
PB[Indigo]
GN[Gnugo 3.0.0 (W) vs. Indigo (B)]
DT[2001-09-06]
SY[Cgoban 1.9.10]TM[30:00(5x31:00)]
RE[B+3.5]
C[This game was played between GNU Go and Indigo on the computer go 9x9
ladder.
Result: Indigo wins by 3.5 due to passive play throughout the game by GNU Go
and a tactical mistake in the endgame.
Comments for GNU Go are made by Inge Wallin, member of the GNU Go team.
Comments for Indigo are made by Bruno Bouzy, author of Indigo.
]
;B[cg]BL[1860];W[df]WL[1860]
C[The first two white moves are taken from GNU Gos opening library.
];
B[gg]BL[1860]
C[Indigo: A move in an empty corner has a very great priority.
];
W[cf]WL[1860];B[cc]BL[1860]
C[Indigo: A move in an empty corner has a very great priority.
];
W[bg]WL[1860]LB[dg:A]
C[GNU Go values this move highly because it takes some territory and has a
high followup value, namely to kill the C3 stone.  The move at D3 (A) was the
second favourite candidate.
]
;B[gc]BL[1860]
C[Indigo: A move in an empty corner has a very great priority.
]
C[Looks a bit thin. Black os open for a splitting attack.
Indigo does not see such splitting attack and still follows the empty corner
occupation rule...
];
W[dg]
WL[1860]
C[GNU Go likes this move because it is kills C3 and is good shape as well. If
white was a professional this would be called honte. :-)
]
;B[ch]BL[1860]
C[Here black starts a futile escape sequence.
Indigo gives a (too) high priority to one-to-three-liberties moves
];W[dh]
WL[1860];B[bh]BL[1860]
C[Indigo does not see that this string is captured.
];
W[bf]WL[1860];B[dd]BL[1860]
C[
Inge Wallin: At this point it is difficult to know what is best.  Black has a
very thin position. But a splitting attack on G5 (A) leads to a complicated
fight with weak groups on both sides (see variation for one example).  A more
careful approach at F5 (B) has bigger chances of succeeding (variation 2).

Indigo now sees that this three-stone string is captured.
Personnally (Bruno Bouzy), I would play F6 (C) but my program does not...
]
LB[ge:A][fe:B][fd:C]
(;W[fh]WL[1860]
C[This is somewhat passive. The proposed move at F5 wasn't even considered.
One of GNU Gos biggest weaknesses is that it doesn't understand the
interaction between weak and strong group and that sometimes attack is the
best defense.
]
;B[gh]BL[1860];W[bd]WL[1860]
C[What follows here are two sente sequences that take some points from black,
but which also leaves black with a big lead.
Indigo answers a little bit passively.]
;B[bc]BL[1860];W[gi]WL[1860];B[hi]BL[1860];W[fi]WL[1860];B[hh]BL[1860]
;W[ac]WL[1860];B[ab]BL[1860];W[ad]WL[1860];B[bb]BL[1860];W[ff]WL[1860]
C[F5 (A) was not considered here.
];B[fg]BL[1860]LB[gf:A]
C[Indigo also considered A];
W[eh]WL[1860]
C[This move looks like a bug (a missing pattern, perhaps).  Both E2 and E1
"defends G1" and "connects D2 and G1" according to GNU Go. E3, the best move
locally here is not on the list.
]
;B[eg]BL[1860]LB[gf:A]
C[Strange. G4 or G5 would win the game for black.
Indigo also considered G4 (A) but this black move is a bug due to a missing
pattern.
];
W[ef]WL[1860];B[gf]BL[1860]
C[
Indigo now plays G4.
]
;W[ed]WL[1860];B[ec]BL[1860]
C[This move is suggested by Indigo's pattern base.
]
;W[fd]WL[1860];
B[fc]BL[1860]
C[This move is suggested by Indigo's pattern base.
]
;W[ge]WL[1860]
C[This is the final game losing move. The correct move is G6 (A).  See
variation.
GNU Go gives this move a very high valuation - 24 points. There seems to be
two bugs here.  One is that there is a too highly valuated followup move to
kill H1. The other one is that it thinks that G5 takes 9 points of territory!
This is probably a problem with the influence module which valuates
territorial values of moves in GNU Go.

With this GnuGo's mistake, Indigo is very lucky.
On this position, Indigo should have played the right move G6 (A).
]
LB[gd:A]
(;B[he]BL[1860];W[gd]WL[1860];B[hd]BL[1860];W[de]WL[1860];B[cd]
BL[1860];W[ce]WL[1860];B[tt]BL[1740];W[tt]WL[1860];
TW[ae][be][ee][fe][af][ag][cg][ah][bh][ch][ai][bi][ci][di][ei]
TB[aa][ba][ca][da][ea][fa][ga][ha][ia][cb][db][eb][fb][gb][hb][ib][dc][hc][ic][id][ie][hf][if][hg][ig][ih][ii]

C[Final score:
   White = 15 territory + 3 captures + 5.5 komi = 23.5
   Black = 27 territory + 0 captures = 27
Black wins by 3.5.
])

(;AE[ge]AW[gd];B[he];W[hd]
(;B[hc];W[ie];B[if];W[id]
(;B[ig];W[ic];B[hb];W[ib];B[ha];W[ge];B[hf]C[White wins
])

(;B[ic];W[ig];B[hf];W[ih]C[A catastrophe for black
])
)

(;B[if];W[hc];B[hb];W[ib];B[ha];W[id];B[ie];W[ge];B[ig]C[White wins
])
)
)

(;W[ge];B[ff];W[fe];B[ee];W[fc];B[gd];W[fd];B[ef])

(;W[fe];B[ge];W[fd];B[gd];W[fc];B[db];W[fb]
(;B[gb];W[ac])

(;B[bd];W[hb])
)

)